1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
Nitrogen fertilisation recommendations : could they be improved using
stochastically generated climates in conjunction with crop models ?
B. Dumont, W. Meza Morales, B. Bodson, and M.-F. Destain
Gembloux Agro-Bio Tech - Department STE & SA
University of Liege
Gembloux, Belgium
B. Basso
Department of Geological Sciences
Michigan State University
Lansing, MI, USA
J.-P. Destain
Phytotechnical Strategy Department
Walloon Agronomical Research Center
Gembloux, Belgium
ABSTRACT
Accurate determination of optimal Nitrogen (N) recommendations which
ensure maximization of farmer's revenue while minimizing the environmental
constraint is maybe among the major challenges in agriculture. Crop models have
the potential to deal with such aspects and could thus be used to develop decision
support systems. However unknown future weather conditions remains the key
point of accurate yield forecast. This paper presents the results of a preliminary
study that aims to supply the unknown future with stochastically generated
climatic conditions. Coupling the methodology with appropriate decision rules led
to a generic decision support system able to guide the N management practices.
Keywords:
Crop model, Strategic N management, Probability risk assessment
INTRODUCTION
In the context of precision nitrogen (N) management, to ensure that the yield
potential could be reached each year, farmers have too often applied quantities of
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
fertilizers much larger than what was strictly required (Lemaire et al., 2008).
However, since 2002, the Belgian Government transposed the European Nitrate
Directive 91/676/EEC in the Belgian law under the Sustainable Nitrogen
Management in Agriculture Program (PGDA) (Vandenberghe et al., 2011), with
the aim to maintain the productivity and the revenue of Belgian's farmers while
reducing the environmental impacts of excessive N management strategies.
At the plot level, crop models are powerful tools to assess the impacts of
different cropping systems inputs, such as agro-environmental conditions (e.g.
soil type, cultivar), management practices (e.g. sowing date, fertilisation), and
climatic conditions (e.g. extreme weather sequences) on the crop harvestable
organs (Dumont et al. 2014a). They are thus particularly suited to support the
decision-making and planning in agriculture (Basso et al., 2011).
However, in a context where increasing frequencies of weather extremes are
observed, much effort have been put in recent years to quantify the climatic
uncertainty impacting the cropping systems productivity. A feasible approach for
evaluating such uncertainty is to quantify the one associated to climatic historical
records (Basso et al., 2011) or to stochastic time-series of weather (Dumont et al.,
2013; Lawless and Semenov, 2005).
Therefore, a promising approach to study the effect of different N fertilisation
practices could be based on the concomitant use of crop models and weather
generators. The objective of this preliminary research was to develop such a
methodology. Its ability to improve N recommendation was then evaluated with
the aim to improve either economic and environmental N-use efficiency.
MATERIALS AND METHODS
Nitrogen fertilisation strategies
In Belgium, the farmers' current N fertilisation practice consists to split a total
180 kgN.ha-1 application into three equal fractions applied respectively at tillering
(Zadoks 23), stem extension (Zadoks 30) and flag-leaf (Zadoks 39) stages.
As long as the final number of grains has not been fixed, the uncertainty linked
to climatic variability impacting the final yields remains very high (Dumont et al.,
2014a). ). However, once the number of grains has been put in place, the end-ofseason yield is driven mainly by the climatic conditions that influence grain
filling, in terms of both carbohydrates and N exportation.
It was therefore decided to impose the first two rates of N applied according to
the current Belgian practice (i.e., 60 kgN.ha-1) applied before anthesis stage. As a
strategic approach, different N levels where then simulated for the flag-leaf stage,
rising from 0 kgN.ha-1 to 100 kgN.ha-1. This strategy will later be referred to as
the ‘Modulation-60 treatment’ (M60-X). In that way, the N protocol
corresponding to the application of 0 kgN.ha-1 would be called M60-0.
Historical records and weather generator
Data originating from a 31-years (1980-2011) weather station were used in this
study. The Ernage station is located only two kilometers from the experimental
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
site. The database provides all the necessary inputs for the crop model (i.e., solar
radiation, wind, precipitation, ambient temperature and relative humidity).
The database was initially analysed using the LARS-Weather Generator
(Semenov and Barrow, 1997). Then, a set of synthetic time-series scenarios
representative of the climatic conditions in the area were stochastically generated.
300 weather time-series were used to ensure stability in predicted mean grain
yield.
16
17
18
19
20
21
22
23
24
25
26
27
28
Eq. 1
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
Decision criteria
The economical optimal N fertilizer is the one that maximise the revenue. The
marginal net revenue (MNR - Eq. 1) can be computed as a function of the benefits
linked to higher yields associated with increasing N practices, and the costs linked
to N applications.
MNR  YN .G P   N .N P 
The grain selling price (Gp) was fixed at 200 eur.ton-1 and the fertilizer price
(Np) was fixed at 0.3 eur.ton-1 with a concentration of 27% of N in the fertilizer.
Furthermore, in the Wallonia region of Belgium, since the PGDA, a survey
system has been developed to control N leaching. Taxes, that take the form of a
reduction of the public financial assistance for farmers', are applied if the total
amount of N remaining in the 0-90 cm soil profile (soil N content [SNC]
kgN.ha-1) is superior to a given threshold. For the purpose of this study, the most
stringent requirements of the directive were considered, i.e. a maximum tax of
120 €.ha-1 is levied if the SNC is superior to 40 kgN.ha-1.
ENR  MNR  Taxes
 0 eur.ha 1 if
Taxes  
1
if
120 eur.ha
Eq. 2
1
SNC  40 kgN.ha
SNC  40 kgN.ha 1
Eq. 3
This criteria will be later referred as the Environmental-friendly Net Revenue
(ENR).
Crop model
The soil-crop model STICS was used in this study. A wide literature can be
found concerning the STICS model formalisms and the way it simulates the yields
(Brisson et al., 2003; Brisson et al., 2009). The STICS model requires daily
weather climatic inputs, namely minimum and maximum temperatures, total
radiation and total rainfall. In the case of more complex formalisms about
potential evapotranspiration calculation, the wind speed and vapour pressure are
needed.
The STICS model parameterisation, involving calibration and validation, was
performed on a four years database presented in the 'Original data' section, using
the DREAM Bayesian parameters sampling algorithm (Dumont et al., 2014b).
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
The contrasted years in terms of climate and yields were used to parameterise the
crop water and thermal stresses dependence.
Original data
Field experiments were carried out to measure the crop response (Triticum
aestivum L.) under temperate climatic conditions (Belgium) in the Hesbaye region
(classic loamy soil). A complete randomised block design was used to evaluate
the crop growth under fertilisation levels ranging from 0 to 240 kgN.ha -1 applied
in two or three fractions. The trials were conducted between 2008 and 2012.
Regular biomass growth references measurements (LAI, total biomass, grain
yield and protein content) as well as soil N content measurements (NO3- and
NH4+) were performed over the growing seasons.
RESULTS AND DISCUSSIONS
Environmental-friendly Net Revenue simulations
Fig. 1. presents the 3D response surface of the environmental-friendly net
revenue simulations. The cumulative distribution functions (CDF) of the ENR
were ranked according to the increasing N practices. The CDF is the immediate
expression of the return time of simulated yields or revenues. As an example,
when 60-60-0 kgN.ha-1 were applied, at least 501.2 eur.ha-1 would be gained 99
years out of 100 (CDF = 0.99). At the same N protocol, they were 50% chances
(1 year out of 2) to achieve at least a revenue of 1807 eur.ha-1. At the highest N
practice, 99 years out of 100 and 1 year out of 2, the revenues would respectively
decrease to 346 eur.ha-1 and increase by 210 eur.ha-1 (2017 eur.ha-1) in
comparison with the lowest N practice.
Under the worst climatic conditions, two different effects could impact the
revenues. On the one hand, the plant was not properly developed to use correctly
high level of N that could be fertilised at flag-leaf stage. Increasing the N practice
led thus to increase the costs without improving the benefits (Eq. 1). On the other
hand, provided the weather would be characterised by excessive rainfall, the
chemical fertilizers would be leached, leading to environmental taxes at the end of
the season.
On Fig. 1, for each probability level, the optimal N practice was highlighted by
a black dot (•). Our observations showed that the optimal N practices seemed
linearly linked to the return time of climatic conditions. Under unfavourable
climatic conditions, the optimal N practices, i.e. the practices that maximised the
ENR, were the lower ones.
According to Basso et al. (2012), a suitable solution would be to select the N
application rate that outperforms the others around 75% of the time. This
threshold was highlighted by a bold dash line (--) on Fig. 1. At this expected
return time, the corresponding optimal practice was the M60-4 protocol, where
40kgN.ha-1 are applied at the flag-leaf stage.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Environmental-friendly net revenue (ENR) as a function of N
fertilisation management and cumulative probability density function (CDF)
drawn from 300 synthetic climate scenarios. The dash line (--) represents the
ENR reached at a probability level of 75% (3 years out of 4). The dots (•)
represents the N treatment producing the optimal ENR under different
probability levels (1%, 5%, 10%, ..., 95%, 99%).
Fig. 1.
Finally, a Wilcoxon test was applied to inter-compare the distributions. The test
revealed that the ENR distribution achieved under the M60-4 protocol were
statistically equivalent to any other superior fertilisation levels (M60-5 to M6010). In the same way, the M60-4 protocol was also equivalent to a practice that
would consist to apply only 20 kgN.ha-1 at flag-leaf. This practice should thus be
preferred as it decreases the environmental risk.
Table 1. Comparison of ENR distribution values using a Wilcoxon test. The
table presents the p-value of the test and the significance level ('*','**','***').
Treatment
M60-9
M60-8
M60-7
M60-6
M60-5
M60-4
M60-3
M60-2
M60-1
M60-0
19
20
M60-10
0.997
0.880
0.669
0.474
0.240
0.098
0.021*
0.002**
0.000***
0.000***
M60-9
M60-8
M60-7
M60-6
M60-5
M60-4
M60-3
M60-2
0.851
0.673
0.430
0.226
0.072
0.016*
0.001**
0.000***
0.000***
0.784
0.553
0.280
0.106
0.020*
0.002**
0.000***
0.000***
0.704
0.395
0.148
0.033*
0.003**
0.000***
0.000***
0.592
0.278
0.067
0.007**
0.000***
0.000***
0.521
0.186
0.022*
0.001***
0.000***
0.440
0.091
0.003**
0.000***
0.317
0.026*
0.191
0.000*** 0.003**
M60-1
0.082
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
CONCLUSION
A methodology was proposed to analyse the impact on crop yields and
farmers' revenues of different N practices under a wide variety of local climate
conditions. Overall the global methodology appeared as a full tool to optimise the
N practice in a strategic point of view.
In particular, our investigation showed that 3 years out of 4, the N fertilizer
rate applied at flag leaf could be decreased by 20 to 40 kgN.ha-1 in comparison to
the current Belgian farmer practice (60-60-60 kgN.ha-1).
In conclusion, the potential of using a crop model as a decision tool for
improving the economical or environmental N-use efficiency was again
demonstrated (Basso et al., 2012). The methodology has furthermore a real
potential to stand as a basis to develop tactical decision support system able to
optimise N practices in real-time
ACKNOWLEDGEMENT
The authors would like to thanks the SPW-DGO3 for its financial support.
They are also thankful to the MACSUR knowledge hub.
REFERENCES
Basso, B., Fiorentino, C., Cammarano, D., Cafiero, G., Dardanelli, J., 2012a.
Analysis of rainfall distribution on spatial and temporal patterns of wheat yield
in Mediterranean environment. Eur. J. Agron. 41(0), 52-65.
Basso, B., Ritchie, J.T., Cammarano, D., Sartori, L., 2011. A strategic and tactical
management approach to select optimal N fertilizer rates for wheat in a spatially
variable field. Eur. J. Agron. 35(4), 215-222.
Brisson, N., Gary, C., Justes, E., Roche, R., Mary, B., Ripoche, D., Zimmer, D.,
Sierra, J., Bertuzzi, P., Burger, P., Bussière, F., Cabidoche, Y.M., Cellier, P.,
Debaeke, P., Gaudillère, J.P., Hénault, C., Maraux, F., Seguin, B., Sinoquet, H.,
2003. An overview of the crop model stics. Eur. J. Agron. 18(3–4) 309-332.
Brisson, N., Launay, M., Mary, B., Beaudoin, N., 2009. Conceptual basis,
formalisations and parameterization of the STICS crop model. Editions Quae.
Collection Update Sciences and technologies.
Dumont, B., Basso, B., Leemans, V., Bodson, B., Destain, J., Destain, M., 2013.
Yield variability linked to climate uncertainty and nitrogen fertilisation. In J.
Stafford (Ed.), Precision agriculture ’13. Wageningen Academic Publishers.
427-434.
Dumont, B., Leemans, V., Ferrandis, S., Vancutsem, F., Bodson, B., Destain, J.,
Destain, M., 2014a. Assessing the potential of predicting wheat yield using a
daily mean climate database. Precis. Agric. 15(3),255-272.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Dumont, B., Leemans, V., Mansouri, M., Bodson, B., Destain, J., Destain, M.,
2014b. Parameter optimisation of the STICS crop model, with an accelerated
formal MCMC approach. Environ. Model. Softw. 52, 121-135
Lemaire, G., Jeuffroy, M.H., Gastal, F., 2008. Diagnosis tool for plant and crop N
status in vegetative stage. Theory and practices for crop N management. Eur. J.
Agron. 28(4), 614-624.
Lawless, C., Semenov, M.A., 2005. Assessing lead-time for predicting wheat
growth using a crop simulation model. Agric. For. Meteorol. 135(1–4) 302-313.
Semenov, M.A., Barrow, E.M., 1997. Use of a stochastic weather generator in the
development of climate change scenarios. Clim. Change 35(4), 397-414.
Vandenberghe, C., Marcoen, J., Sohier, C., Degre, A., Hendrickx, C., Paulus, P.,
2011. Monitoring networks and modelling systems for assessing effectiveness
of the EU Nitrates Directive Action Programmes: Approach by the Walloon
Region (Belgium). Results of the second International Workshop, 10-11 June
2009.